You are on page 1of 80

HST.

725 Music Perception and Cognition, Spring 2009


Harvard-MIT Division of Health Sciences and Technology
Course Director: Dr. Peter Cariani

HST 725

Music Perception & Cognition

Lecture 3

What we hear:

Basic dimensions of auditory experience

www.cariani.com

Tuesday, February 10, 2009


What we hear: dimensions of auditory experience
Hearing: ecological functions (distant warning,
communication, prey detection; works in
the dark)
Detection, discrimination, recognition, reliability, scene analysis
Operating range: thresholds, ceilings, & frequency limits
Independent dimensions of hearing & general properties
Pitch
Timbre (sound quality)
Loudness
Duration
Location
Distance and Size
Perception of isolated pure tones
Interactions of sounds: beatings, maskings, fusions
Masking (tones vs. tones, tones in noise)
Fusion of sounds & the auditory "scene":
how many objects/sources/voices/streams?
Representation of periodicity and spectrum
Tuesday, February 10, 2009
MECHANISM Sensory
encodings
Receptors External
Neural codes Effectors world
Motor
commands

Neural
architectures

Information-processing
operations

Functions

Tuesday, February 10, 2009


Hearing: ecological functions

Distant warning of predators


approaching
Identication of predators
Localization/tracking of prey
Con-specic communication
Mating/competition
Cooperation (info. sharing)
http://www.batsnorthwest.org/

Territory Scott Pederson.


Corynorhinus townsendii -
Townsend's Big-eared Bat

Navigation in the dark


General recognition of sounds

http://www.pbs.org/wgbh/nova/wolves/

http://www.pbs.org/lifeofbirds/songs/index.html

Photo: US Fish and Wildlife Service.

Tuesday, February 10, 2009


The Effect of Context on Information Processing

in the Auditory System

Ell

Ellen Covey

University of Washington

Seattle, WA Courtesy of Ellen Covey. Used with permission.

Tuesday, February 10, 2009


Jim Simmons, Brown

Courtesy of James Simmons. Used with permission.

Tuesday, February 10, 2009


The auditory scene:
basic dimensions

Temporal organization
Events
Notes
Temporal patterns of events

Organization of sounds
Voices, instruments
Streams
Objects Paul Cezanne. "Apples, Peaches, Pears and Grapes."
Sources Courtesy of the iBilio.org WebMuseum.


Attributes of sounds
Loudness (intensity)
Pitch (dominant periodicity)
Timbre (spectrum)
Duration
Location (bearing, range)
Tuesday, February 10, 2009
Auditory qualities in music perception & cognition

Pitch Melody, harmony, consonance


Timbre Instrument voices
Loudness Dynamics
Organization Fusions, objects. How many voices?

Rhythm Temporal organization of events


Longer pattern Repetition, sequence

Mnemonics Familiarity, novelty


Hedonics Pleasant/unpleasant
Affect Emotional associations, meanings
Semantics Cognitive associations/expectations

Tuesday, February 10, 2009


Dimensions of auditory objects Dimensions of event perception

Auditory qualities and their organization Unitary events & their organization

Objects: Quasi-stationary Events: abrupt perceptual


assemblages of qualities discontinuities

TEMPORAL
Duration
EVENT
Pitch STRUCTURE
Loudness
Timing & order
Timbre Location (metric, sequence)
Spatial Dimensions

FUSION/SEPARATION FUSION/SEPARATION
Common onset & harmonic structure => fusion Common onset, offset => fusion
Different F0s, locations, onset => separation Diff. meters, pitch, timbre => separation
POLYPHONY STREAMS, POLYRHYTHMS

Tuesday, February 10, 2009


Visual scene

Line

Shape

Texture

Lightness

Color

Transparency

Objects
Photo removed due to copryight restrictions. Cover of LIFE Magazine, Nov. 23, 1936.
View in Google Books.
Apparent distance

Apparent size

etc.

Margaret Bourke-White
Fort Peck Dam 1936

Tuesday, February 10, 2009


Psychophysics
Stimulus Perceptual
performance

Characterizing
Reverse

& predicting
engineering

neural responses

How does it work?

Why does the system


What aspects of

respond as it does?
neural response

are critical for perceptual

Stimulus
function?

Retrodiction
Neural
Retrodiction of percepts

Transmission
responses
Correspondence

delity
Between putative

Estimates of
neural representation

channel capacity
& perceptual performance

Biophysical Psychoneural
models neurocomputational models
Tuesday, February 10, 2009
Perceptual functions
Subjective vs. objective measures

Subjective measures

Magnitude estimation

Objective measures

Detection: capability of distinguishing the presence or absence of a stimulus

(or some aspect of a stimulus, e.g. AM detection)

Threshold: the value of a stimulus parameter at which a stimulus can be reliably

detected

Sensation level (SL): sound level re: threshold

Discrimination: capability of distinguishing between two stimuli

Difference limen: the change in a stimulus parameter required for reliable

discrimination, just-noticeable-difference (jnd)

Weber fraction: Difference limen expressed as proportional change (e.g. f/f)

Matching task: subject changes parameter that matches two stimuli

Two-alternative forced choice (2AFC)

Ranking tasks

Recognition: correct identication of a particular stimulus

Masking: impairment of ability to detect a signal in the presence of other signals

Tuesday, February 10, 2009


Vibrations create compressions and expansions of air

Sound waves are alternating local


changes in pressure
These changes propagate through
space as longitudinal waves

Condensation phase (compression):

pressure increases

Rarefaction (expansion) phase:

pressure decreases

Source: Handel, S. Listening: an Introduction to the Perception of Auditory Events. Handel


Tuesday, February 10, 2009 Cambridge, MA: MIT Press, 1989. Courtesy of MIT Press. Used with permission.
Waveforms
Microphones convert sound pressures to electrical voltages.

Waveforms plot pressure as a function of time, i.e. a time-series of amplitudes.

Waveforms are complete descriptions of sounds.

Audio CDs sample sounds at 44,100 samples/sec.

Oscilloscope demonstration.

Figure by MIT OpenCourseWare.

Tuesday, February 10, 2009


Oscilloscope demonstration
Waveforms plot pressure as a function of time, i.e. a time-series of amplitudes.
Waveforms are complete descriptions of sounds.

Tuesday, February 10, 2009


Sampling rate (samples/second)
Nowadays sounds are usually converted to strings of numbers that indicate
sound pressure or voltage at each of many equally spaced points in time. THe
number of samples collected per sec is the sampling rate.

Tuesday, February 10, 2009


From sound to numbers
CD quality sound
sound pressure changes 16 bits = 216 = 64k
voltage levels,
microphone sampling rate @
44,100 samples/sec
electrical voltage changes

digitizer

(analog to digital converter)

numerical values, time-series

The upper limit of human hearing is ~ 20,000 cycles/sec (Hertz, Hz).

There is a theorem in signal processing mathematics that the highest frequency that can

be represented is 1/2 the sampling rate (called the Nyquist frequency).

This is why sound for CDs is sampled at 44.1 kHz.

In theory, this is the point where all sound distinctiona we can hear is captured.

MP3s compress the description by about 10-fold (we will discuss later).

Tuesday, February 10, 2009


Sound level basics
Sound pressure levels are measured relative to an absolute
reference
(re: 20 micro-Pascals, denoted Sound Pressure Level or SPL).

Since the instantaneous sound pressure uctuates, the


average amplitude of the pressure waveform is measured
using root-mean-square RMS. (Moore, pp. 9-12)
Rms(x) = sqrt(mean(sum(xt2)))
Where xt is the amplitude of the waveform at each instant t in the
sample
Because the dynamic range of audible sound is so great, magnitudes
are expressed in a logarithmic scale, decibels (dB).
A decibel of amplitude expresses the ratio of two amplitudes
(rms pressures, P1 and P_reference) and is given by the
equation:
dB = 20 * log10(P1/P_reference)
20 dB = 10 fold change in rms level
Tuesday, February 10, 2009
Decibel scale for relative amplitudes (levels)
(rules of thumb)

20 dB = fold change amplitude


10 dB = 3+ fold change
6 dB = 2 fold change amplitude
3 dB = 1.4 fold change
2 dB = 1.26 fold change (26 %)
1 dB = 1.12 fold change (12%)
0 dB = 1 fold change (no change)
-6 dB = 1/2
-20 dB = 1/10 fold change

Tuesday, February 10, 2009


Dynamic range

0 dB SPL is set at 20 microPascals


60 dB SPL is therefore a 1000 fold change in RMS over 0 dB

A typical background sound level is 50-60 dB SPL.

Dynamic range describes the range of sound pressure levels.

The auditory system registers sounds from 20 dB to >> 120 dB SPL

The auditory system has a dynamic range in excess of 100 dB (!) or a


factor of 105 = 100,000 in amplitude.

It is quite remarkable that musical sounds remain recognizable over


most of this range. This a fundamental aspect of hearing that all
auditory theories must address -- how auditory percepts remain
largely invariant over this huge range (perceptual constancy).
Tuesday, February 10, 2009
Hearing has a huge dynamic range!

Hearing has a huge dynamic range!

The dynamic range of human


hearing is the ratio of the sound CD quality sound
pressure level of the softest sound that 16 bits = 216 = 64k
can be heard to the loudest one that voltage levels,
can be tolerated without pain. sampling rate @
44,100 samples/sec
This dynamic range is > 100,000 (>
100 dB or 105 fold), and is roughly
comparable to the 65,536 amplitude
steps that are afforded by 16-bit
digitization.

Tuesday, February 10, 2009


Typical sound levels in music On origins of music dynamics notation
http://www.wikipedia.org/wiki/Pianissimo

Text removed due to copyright


restrictions. See the Wikipedia
article.

Pain > 130 dB SPL


Loud rock concert 120 dB SPL
Loud disco 110 dB SPL
fff 100 dB

SPL

f (forte, strong) 80 dB

SPL

p (piano, soft) 60 dB

SPL

ppp 40 dB SPL

Lower limit
Theshold of hearing 0 dB SPL
Tuesday, February 10, 2009
Typical sound pressure levels in everyday life

Disco

Courtesy of WorkSafe, Department of Consumer and Employment Protection, Western Australia (http://www.safetyline.wa.gov.au).

Tuesday, February 10, 2009


Demonstrations

Demonstrations using waveform generator

Relative invariance of pitch & timbre with level

Loudness matching
Pure tone frequency limits
Localization

Tuesday, February 10, 2009


Loudness

Dimension of perception that changes with

sound intensity (level)

Intensity ~ power;

Level~amplitude

Demonstration using waveform generator

Masking demonstrations

Magnitude estimation

Loudness matching

Tuesday, February 10, 2009


Sound level meters and frequency weightings

A-weighting: perceived loudness

C-weighting: at

5
0 A
-5 C
Relative Response (dB)

B, C
-10
-15
-20 B
-25
-30
-35
A
-40
-45
2 4 8 102 2 4 8 103 2 4 8 104

Frequency (Hz)

Graph by MIT OpenCourseWare. SPL meter photo courtesy of EpicFireworks on Flickr.

Tuesday, February 10, 2009


Loudness as a function of pure tone level & frequency

Absolute detection thresholds


on the order of
1 part in a million,
pressure ~1/1,000,000 atm
(Troland, 1929)

Figure by MIT OpenCourseWare.

Tuesday, February 10, 2009


Loudness perception:

Tuesday, February 10, 2009 Figure by MIT OpenCourseWare.


Loudness perception: population percentiles

Figure by MIT OpenCourseWare.

Tuesday, February 10, 2009


Intensity discrimination improves at higher sound levels

Best Weber fraction

Figure by MIT OpenCourseWare.

Tuesday, February 10, 2009


Hearing loss with age

Tuesday, February 10, 2009 Figure by MIT OpenCourseWare.


Dynamic range of some
musical instruments

Images removed due to copyright restrictions.

Graphs of relative intensity vs. pitch for different instruments: violin, double bass, flute, B-flat clarinet, trumpet, french horn.

Figure 8.5 in Pierce, J. R. The Science of Musical Sound. Revised ed. New York, NY: W.H. Freeman & Co., 1992. ISBN: 9780716760054.

Tuesday, February 10, 2009


Periodicity and spectrum
Periodicity vs. frequency
Longstanding and ongoing dichotomy between
formally-equivalent, yet complementary
perspectives (de Cheveigne chapter on
pitch)

Vibrating strings vs. Helmholtz resonators

Comb filters vs. band-limited filters

Autocorrelation vs. Fourier analysis

and yet another paradigm


Complex oscillator (delay loop)
Tuesday, February 10, 2009
Complex modes of vibration

Most physical systems have multiple

modes of vibrations that create

resonances that favor particular

sets of frequencies.

Vibrating strings or vibrating columns

of air in enclosures exhibit

harmonic resonance patterns.

Material structures that are struck

(bells, xylophones, percussive

instruments) have resonances

that depend partly on their

shape and therefore

canproduce frequencies that

are not harmonically related.

More later on what this means for Figure by MIT OpenCourseWare.

pitch and sound quality.

Tuesday, February 10, 2009


Frequency spectra
The Greeks understood simple relationships between vibration rate & pitch.
Experiments with musical instruments and tuning systems were carried out by
many people (Galileo s father, Galileo, Saveur, Mersenne, others).

Joseph Fourier (1768-1830) showed that any waveform can be


represented as the sum of many sinusoids ( Fourier spectrum).

George Ohm (1789-1854) postulated that sounds can be decomposed into component sinusoids
Hermann von Helmholtz (1821-1894) postulated that the ear analyzes sound
by rst breaking sounds into their partials and then doing associative pattern-recognition
Debate between Seebeck, Ohm, & Helmholtz (1844) over periodicity vs. spectral pattern
Foreshadows temporal vs. place codes, autocorrelation vs. Fourier spectrum

Each sinusoid of a particular frequency (frequency component, partial) has 2 parameters:


1) its magnitude (amplitude of the sinusoid)
2) its phase (relative starting time)

A sound with 1 frequency component is called a pure tone.


A sound with more than one is called a complex tone.

Tuesday, February 10, 2009


Fundamentals and harmonics

Periodic sounds (30-20kHz)

produce pitch sensations.

Periodic sounds consist of

repeating time patterns.

The fundamental period (F0) is the

duration of the repeated pattern.

The fundamental frequency is the

repetition frequency of the pattern.

In the Fourier domain, the

frequency components of a

periodic sound are all members of

a harmonic series (n = 1*F0, 2*F0,

3*F0...).

The fundamental frequency is

therefore the greatest common

divisor of all of the component

frequencies.

The fundamental is also therefore

a subharmonic of all component

frequencies.

Tuesday, February 10, 2009 Figure by MIT OpenCourseWare.


Harmonic series

A harmonic series consists of integer multiples of a fundamental frequency,


e.g. if the fundamental is 100 Hz, then the harmonic series is: 100, 200,
300, 400, 500, 600 Hz, .... etc.
The 100 Hz fundamental is the rst harmonic, 200 Hz is the second

harmonic. The fundamental is often denoted by F0.

The fundamental frequency is therefore the greatest common divisor of all


the frequencies of the partials.

Harmonics above the fundamental constitute the overtone series.

Subharmonics are integer divisions of the fundamental:

e.g. for F0= 100 Hz, subharmonics are at 50, 33, 25, 20, 16.6 Hz etc.

Subharmonics are also called undertones.

The fundamental period is 1/F0, e.g. for F0=100 Hz, it is 1/100 sec or 10

Tuesday, February 10, 2009


Sound quality contrasts

Impulsive sounds
Duration
Sustained sounds
Stationary vs. nonstationary

Pitched sounds
Time domain: Periodic sound patterns
Pattern Frequency domain: harmonics
complexity, Inharmonic sounds
coherence Combinations of unrelated periodic patterns
Complexity: Number of independent patterns
Noises
Aperiodic sound patterns, high complexity

Tuesday, February 10, 2009


Minimal

durations

Graph removed due to copyright restrictions.

Figure 36, comparing "Tone pitch" and "click pitch" responses.

In Licklider, J. C. R. "Basic Correlates of the Auditory Stimulus."

Handbook of Experimental Psychology. Edited by S. S. Stevens.

Oxford, UK: Wiley, 1951. pp. 985-1039.

Licklider (1951)

Basic correlates

of the auditory stimulus

Tuesday, February 10, 2009


Periodic vs. aperiodic sounds
Periodic sound patterns -- tones

Aperiodic sound patterns -- noise

Tuesday, February 10, 2009


Range of pitches of pure & complex tones

Pure tone pitches


Range of hearing (~20-20,000 Hz)
Range in tonal music (100-4000 Hz)

Most (tonal) musical instruments


produce harmonic complexes that
evoke pitches at their fundamental
frequencies (F0s)
Range of F0s in tonal music (30-4000 Hz)
Range of missing fundamental (30-1200 Hz)

Tuesday, February 10, 2009


JND's

Figure by MIT OpenCourseWare.

Tuesday, February 10, 2009


Pure tone pitch discrimination
becomes markedly worse
above 2 kHz

Weber fractions for


frequency (f/f) increase
1-2 orders of magnitude
between 2 kHz and 10 kHz

Figure by MIT OpenCourseWare.

Tuesday, February 10, 2009


Pure tone
pitch
discrimination
improves

at longer
tone
durations

and

at
higher
sound
pressure
levels

Figure by MIT OpenCourseWare.

Tuesday, February 10, 2009


Emergent pitch

Missing Line spectra Autocorrelation (positive part)


F0
600

10

0 5 10 15 20

0 200 400 600 800 1000 1200 1400 1600 1800 2000

Pure tone

200 Hz
0 5 10 15 20

0 200 400 600 800 1000 1200 1400 1600 1800 2000

Tuesday, February 10, 2009


Correlograms: interval-place displays (Slaney & Lyon)

Frequency (CF)

Frequency (CF)

Autocorrelation lag (time delay)


Courtesy of Malcolm Slaney (Research Staff Member of IBM Corporation). Used with permission.

Tuesday, February 10, 2009


8k

Frequency ranges of (tonal) musical instruments 6

> 6 kHz 2.5-4 kHz 3

27 Hz 110 262 440 880 4 kHz

Hz Hz Hz Hz

Tuesday, February 10, 2009


Frequency ranges: hearing vs. musical tonality

100 Hz 2 kHz

Temporal
neural
mechanism

Musical tonality
Octaves, intervals, melody: 30-4000 Hz

Range of hearing Place


Ability to detect sounds: ~ 20-20,000 Hz mechanism
Tuesday, February 10, 2009
Duplex time-place representations

temporal representation
level-invariant
strong (low fc, low n) place-based
weak (high fc, high n; F0 < 100 Hz) representation
level-dependent
coarse

30 100 1k 10k

Similarity
to cf. Terhardt's
interval spectral and virtual pitch
pattern

Similarity to place pattern

Tuesday, February 10, 2009


A "two-mechanism" perspective (popular with some psychophysicists)

harmonic number

unresolved harmonics
weak temporal mechanism
phase-dependent; rst-order intervals
n= 5-10

place-based place-based

resolved harmonics representations

representation
strong spectral pattern mechanism
level-independent
phase-independent level-dependent
rate-place? interval-place?
ne coarse

30 100 1k 10k

f, F0 Dominance region

Tuesday, February 10, 2009


Pitch dimensions: height & chroma

Figure by MIT OpenCourseWare.

Tuesday, February 10, 2009


Pitch height and pitch chroma

Images removed due to copyright restrictions.


Figures 1, 2, and 7 in Shepard, R. N. "Geometrical approximations to the structure of musical pitch."
Psychological Review 89, no. 4 (1982): 305-322.

Tuesday, February 10, 2009


Two codes for pitch

Place code Time code

Pitch = place of excitation Pitch = dominant periodicity

Pitch height
Musical pitch
"Chroma", Tonality

Absolute Relational

Low vs. high Musical intervals, tonality


Melodic recognition
& transposition
Existence region
Existence region
f: .100-20,000 Hz
F0: 30-4000 Hz

Tuesday, February 10, 2009


Note durations in music

Figure by MIT OpenCourseWare.

Tuesday, February 10, 2009


Timbre: a multidimensional tonal quality

(Photo Courtesy of Pam Roth.


tone texture, tone color
Used with permission.)
distinguishes voices,

instruments

Stationary Dynamic

Aspects Aspects

Photo Courtesy of Per-Ake


(spectrum) spectrum Bystrom.Used with permission.)

intensity
pitch
Vowels attack
decay
Photo Courtesy of Miriam
Lewis. Used with permission.)
Consonants
Stationary spectral aspects of timbre

Waveforms Power Spectra Autocorrelations


Formant-related Pitch periods, 1/F0
Vowel quality 125 Hz 100 Hz
Timbre

[ae]
F0 = 100 Hz

[ae]
F0 = 125 Hz

[er]
F0 = 100 Hz

[er]
F0 = 125 Hz

0 10 20 0 1 2 3 4 0 5 10 15
Time (ms) Frequency (kHz) Interval (ms)
Tuesday, February 10, 2009
Rafael A. Irizarry's Music and Statistics Demo

Spectrograms
of Harmonic Instruments

violin, trumpet, guitar

Courtesy of Rafael A. Irizarry. Used with permission.

Non-Harmonic Instruments

marimba, timpani, gong

Audio samples at this page:


http://www.biostat.jhsph.edu/~ririzarr/Demo/demo.html

Tuesday, February 10, 2009


Timbre dimensions: spectrum, attack, decay

Courtesy of Hans-Christoph Steiner.


Used with permission. After J. M. Grey,
Stanford PhD Thesis (1975) and Grey and
Gordon, JASA (1978).

Tuesday, February 10, 2009


Interference interactions between tones

[Public domain image]

Tuesday, February 10, 2009


Masking audiograms

[Public domain image]

Wegel & Lane, 1924

Tuesday, February 10, 2009


1000 Hz pure tone masker

Graph removed due to copyright restrictions.


See Fig. 3, "Average masking patterns for 1000 cps based upon three listeners" in Ehmer,
Richard H. "Masking Patterns of Tones." The Journal of the Acoustical Society of America,
vol. 31, no. 8 (1959): 1115. http://www.zainea.com/masking2.htm

Tuesday, February 10, 2009


Tone on tone masking curves (Wegel & Lane, 1924)

[Public domain images]

Tuesday, February 10, 2009


Resolvability of harmonics (Plomp, 1976)

Image removed due to copyright restrictions.


Graph of frequency separation between partials vs. frequency of the partial. From Plomp, R. Aspects of Tone Sensation.
New York, NY: Academic Press, 1976.

Tuesday, February 10, 2009


From masking patterns
to "auditory lters" as a
model of hearing

Power spectrum

Filter metaphor

Notion of one central


spectrum that subserves

2.2. Excitation pattern. Using the lter shapes and bandwidths derived from masking experiments we can
produce the excitation pattern produced by a sound. The excitation pattern shows how much energy comes
through each lter in a bank of auditory lters. It is analogous to the pattern of vibration on the basilar
membrane. For a 1000 Hz pure tone the excitation pattern for a normal and for a SNHL (sensori-neural
hearing loss) listener look like this: The excitation pattern to a complex tone is simply the sum of the
patterns to the sine waves that make up the complex tone (since the model is a linear one). We can hear out
a tone at a particular frequency in a mixture if there is a clear peak in the excitation pattern at that
frequency. Since people suffering from SNHL have broader auditory lters their excitation patterns do not
have such clear peaks. Sounds mask each other more, and so they have difculty hearing sounds (such as
speech) in noise. --Chris Darwin, U. Sussex

Courtesy of Prof. Chris Darwin (Dept. of Psychology at the University of Sussex).

Tuesday, February 10, 2009


Shapes of perceptually-derived "auditory lters" (Moore)

Dont conate these with cochlear lters or auditory


nerve excitation patterns! Auditory lters are derived
from psychophysical data & reect the response of the
whole auditory system. For lower frequencies and higher
levels AFs have much narrower bandwidths than
cochlear resonances or auditory nerve ber responses.

Figures by MIT OpenCourseWare.

Tuesday, February 10, 2009


Resolution of harmonics

Figure by MIT OpenCourseWare.

Tuesday, February 10, 2009


A "two-mechanism" perspective (popular with some psychophysicists)

harmonic number

unresolved harmonics
weak temporal mechanism
phase-dependent; rst-order intervals
n= 5-10

place-based place-based

resolved harmonics representations

representation
strong spectral pattern mechanism
level-independent
phase-independent level-dependent
rate-place? interval-place?
ne coarse

30 100 1k 10k

f, F0 Dominance region

Tuesday, February 10, 2009


.

Stimulus

Spectral pattern
50 F0= 80 Hz

analysis
dB

vs.
0 Frequency (kHz) 3

array of cochlear band-pass filters

temporal pattern
auditory
analysis
nerve fiber
tuning curves
discharge rates interspike intervals
Note: Some models,
such as Goldstein's
Power spectrum
Autocorrelation

use interspike interval representation


representation

information to rst form Frequency domain Time domain


a Central Spectrum Population rate-

which is then analyzed using place profile


Population
interspike interval
harmonic spectral templates. distribution
1/F0
There are thus dichotomies

# intervals

optimal frequency
1/F1
1) between use of match
(linear scale)

time and placeinformation


F0 = 200 Hz
as the basis of the central
representation, and F0 = 160 Hz 0 5 10 15
F0 = 100 Hz Interval (ms)
2) use of spectral vs.
autocorrelation-like central harmonic templates
representations
Pitch best fitting template

Tuesday, February 10, 2009


JNDs for pure tone frequency (Roederer, 1995)

Graph removed due to copyright restrictions.


Fig. 2.9 in Roederer, J. G. The Physics and Psychophysics of Music: An Introduction.
New York, NY: Springer, 1995.

Tuesday, February 10, 2009


Critical bands Partial fusion,
as tonal fusion partial loudness summation

(Roederer, 1995)

Two sine waves, Graph removed due to copyright restrictions.


one fixed at 400 Hz, Fig. 2.12 in Roederer, J. G. The Physics and Psychophysics of Music: An Introduction.
the other ascending New York, NY: Springer, 1995.
from 400 Hz to 510 Hz See 2nd graph on this page:

at which point it is http://www.sfu.ca/sonic-studio/handbook/Critical_Band.html

separated from the


first by a
critical bandwidth.

Tuesday, February 10, 2009


Critical
bandwidths
(Roederer, 1995)

Graph removed due to copyright restrictions.


Fig. 2.13 in Roederer, J. G. The Physics and Psychophysics of Music: An Introduction.
New York, NY: Springer, 1995.
See 1st graph on this page: http://www.sfu.ca/sonic-studio/handbook/Critical_Band.html

Tuesday, February 10, 2009


Critical bands (conventional, place-based interpretation)

CRITICAL BAND and CRITICAL BANDWIDTH


For a given FREQUENCY, the critical band is the smallest BAND of frequencies around it
which activate the same part of the BASILAR MEMBRANE. Whereas the DIFFERENTIAL
THRESHOLD is the just noticeable difference (jnd) of a single frequency, the critical
bandwidth represents the ear's resolving power for simultaneous tones or partials.

In a COMPLEX TONE, the critical bandwidth corresponds to the smallest frequency


difference between two PARTIALs such that each can still be heard separately
also be measured by taking a SINE TONE barely MASKed by a band of WHITE NOISE
around it; when the noise band is narrowed until the point where the sine tone
. It may
becomes audible, its width at that point is the critical bandwidth. See: RESIDUE

In terms of length (see diagram under BASILAR MEMBRANE) the critical bandwidth is .nearly
constant at 1.2 mm, within which are located about 1300 receptor cells, and is
generally independent of intensity (unlike COMBINATION TONES). Twenty-four critical
bands of about one-third octave each comprise the audible spectrum.
Truax, B., ed. From "CRITICAL BAND and CRITICAL BANDWIDTH." http://www.sfu.ca/sonic-studio/handbook/Critical_Band.html
Handbook for Acoustic Ecology. 2nd edition, 1999. Courtesy of Barry Truax. Used with permission.

Tuesday, February 10, 2009


Critical bands (usually interpreted in terms of frequency analysis)

Simultaneous tones lying within a critical bandwidth do not give any increase in

perceived loudness over that of the single tone, provided the sound pressure level

remains constant. For tones lying more than a critical bandwidth apart, their

combination results in increased loudness.

When two tones are close together in frequency, BEATS occur, and the resulting
tone is a fusion of the two frequencies. As the frequency difference increases,
roughness in the tones appears, indicating that both frequencies are activating the
same part of the basilar membrane. Further apart, the two frequencies can be
discriminated separately, as shown below by DfD, whereas roughness only
disappears at a frequency separation equal to the critical bandwidth DfCB. At this
point, the two frequencies activate different sections of the basilar membrane. This
phenomenon only applies to monaural listening with pure tones. With DICHOTIC
listening, the basilar membrane of each ear is activated separately, and therefore no
roughness results. With complex tones, frequency discrimination is improved but the
critical bandwidth remains the same for each of the component partials.

Alternative interpretation is that critical bandwidths are the result of fusion of


(e.g. interspike interval) representations rather than cochlear proximity per se.
Truax, B., ed. From "CRITICAL BAND and CRITICAL BANDWIDTH." http://www.sfu.ca/sonic-studio/handbook/Critical_Band.html
Handbook for Acoustic Ecology. 2nd edition, 1999. Courtesy of Barry Truax. Used with permission.

Tuesday, February 10, 2009


Masking by signal swamping
Reduce signal/noise to disrupt
signal detection

Camouage: pattern fusion


Disruption of pattern detection

Re: varieties of masking in the auditory


system, see Delgutte (1988)
Physiological mechanisms of masking.
In. Duifhuis, Horst & Wit, eds. Basic
Issues in Hearing. London. Academic
Photo courtesy of kirklandj on Flickr.
Press, 204-14.
Tuesday, February 10, 2009
Spatial hearing

Azimuth:
interaural time differences (20-600 usec)
interaural level differences

Elevation:
received spectrum of broadband sounds (pinna effects)

Distance
Spatial form (size, shape)
Enclosure size, shape Image removed due to copyright restrictions.
Diagram showing effect of interaural path-length differences.
Reverberation pattern Figure 2.1 in Warren, R. M. Auditory Perception: A New Synthesis.
New York, NY: Pergamon Press, 1982. ISBN: 9780080259574.
Patterns of long delays

Assignment of spatial attributes


to auditory objects
Tuesday, February 10, 2009
Interaural time difference and localization of sounds

Figure by MIT OpenCourseWare.

Tuesday, February 10, 2009


Binaural cross-correlation and cancellation

Binaurally-created pitches

Tones (F0 from one harmonic in each ear)

Phase-disparity pitches (auditory analog

of Julez random-dot stereodiagrams)

Repetition pitches (weak)

Binaural masking release (BMLD)

A tone in noise that is just masked is

presented to one ear. The tone cannot be heard initially.

Now also present the identical noise alone in the other ear

and the tone pops out. The noise appears to be cancelled

out, providing up to 15 dB of unmasking.

Tuesday, February 10, 2009


Generalist vs. specialist sensory systems (conjectures)

General-purpose vs. special-purpose systems


Adaptability vs. adaptedness
Adaptable: optimized for many different envs

Adaptedness: high degree of optimization

Tradeoff between the two

Panda gut (highly adapted) vs. human gut (omnivore, high adaptability)

In sensory systems, high adaptability is favored when appearances are


highly variable; adaptedness when appearances are highly constrained
Intra-species communications: adaptedness is favored
Signal production and reception under same genetic coordination
e.g. pheromone systems
Inter-species interactions (predator or prey): adaptability is favored under
varying relations, adaptedness under stable relations
(e.g. navigation systems, early warning systems, predator or prey
recognition)
Tuesday, February 10, 2009
Reading for Tuesday, Feb 13

Next meeting we will introduce neural coding and

give an overview of the auditory system.

Weinberger chapter in Deutsch (3)

Look over auditory physiology chapter in Handel (12)

Tuesday, February 10, 2009


MIT OpenCourseWare
http://ocw.mit.edu

HST.725 Music Perception and Cognition


Spring 2009

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.